Risk Bands That Do Not Split Ties
Medium About 15 min Katabench Pro
Assign percentile risk bands with CUME_DIST so accounts with equal scores remain together.
The full brief is part of Katabench Pro
Pro opens this kata's brief, its starter code and its grader. To see how the track works first, try Customer Order Totals, which is free.
How it is graded
Your LINQ or SQL runs against real PostgreSQL. You see the SQL your code generates on the sample tests and can inspect its execution plan, so you learn when the database did the wrong work.
What it practices
- Window functions
- LAG, LEAD, running sums and shares computed over partitions without collapsing rows.
- Statistical aggregates
- Medians, percentiles, distributions, weighted averages and regression computed in SQL.
Related katas
- Average by Time, Not Samples Hard Pro Reconstruct state intervals with the prior sample and LEAD, clip them to the report, and compute a duration-weighted average.
- One Contact per Email Medium Pro Keep one canonical row per duplicate group with ROW_NUMBER over a priority ordering, not GROUP BY aggregates.
- Three Price Tiers of Depth Medium Pro Use a GROUPS window frame to measure rolling auction depth across distinct price tiers without splitting peers.
- Rank a Release Before It Ships Medium Pro Use a hypothetical-set RANK aggregate to place an unpersisted release in every channel leaderboard.